=========================== xia2 multi-crystal tutorial =========================== In this tutorial we are going to process several 5° wedges of thermolysin data, taken from multiple crystals, first using xia2 to process the images using the XDS pipeline, and then use BLEND_ on the resulting processed reflection files to analyse the datasets for non-isomorphism. First we need to type several commands in order to make available all the software we need for the tutorial:: # make XDS available module load XDS # get the latest version of CCP4 module load ccp4 # get latest combined DIALS/xia2 module load dials/dev-228 # needed by BLEND module load R Next we have the xia2 command to run. Below we explain in detail the meaning of each parameter:: xia2 -3dii -failover \ mode=parallel njob=4 nproc=1 \ trust_beam_centre=True read_all_image_headers=False \ unit_cell="92.73 92.73 128.13 90.00 90.00 120.00" \ space_group=P6122 -atom X \ /dls/mx-scratch/mx_bag_training/thermolysin/ This command will run all the processing on a single computer. Assuming the computer has 4 processors available, we have set ``njob=4``. If the computer we are using has more processors available, we may increase either ``njob`` or ``nproc``, whilst ensuring that ``nproc x njob`` is less than or equal to the total number of processors available. Alternatively, we can take advantage of the cluster to process our data more quickly, by using the xia2 parameter ``type=qsub``, which tells xia2 to submit individual jobs to the cluster using the ``qsub`` command. Since we are using the cluster, and not a single computer, we can increase the ``nproc`` parameter, e.g. to ``nproc=12`` which will now use 12 processors per sweep. We can also increase ``njob=20`` to process all 20 sweeps simultaneously. We also need to add an extra ``module load global/cluster`` command to make the cluster available:: module load global/cluster module load XDS module load ccp4 module load dials/dev-228 xia2 -3dii -failover \ mode=parallel njob=20 nproc=12 type=qsub\ trust_beam_centre=True read_all_image_headers=False \ unit_cell="92.73 92.73 128.13 90.00 90.00 120.00" \ space_group=P6122 -atom X \ /dls/mx-scratch/mx_bag_training/thermolysin/ +----------------------------------+-------------------------------------------+ | Parameter | Description | +==================================+===========================================+ | ``-3dii`` | Use the 3D (XDS) pipeline, using all | | | images in indexing | +----------------------------------+-------------------------------------------+ | ``-failover`` | If processing fails for any sweeps, | | | ignore and just use those sweeps that | | | processed successfully | +----------------------------------+-------------------------------------------+ | ``mode=parallel`` | Process multiple sweeps in parallel, | | | rather than serially | +----------------------------------+-------------------------------------------+ | ``njob=4`` | In conjunction with mode=parallel, | | | process 4 sweeps simultaneously | +----------------------------------+-------------------------------------------+ | ``nproc=1`` | Use 1 processor per job (sweep) | +----------------------------------+-------------------------------------------+ | ``type=qsub`` | Submit individual processing jobs to | | | cluster using qsub | +----------------------------------+-------------------------------------------+ | ``trust_beam_centre=True`` | Don't run labelit beam centre search | +----------------------------------+-------------------------------------------+ | ``read_all_image_headers=False`` | Skip reading all image headers - just | | | read the first one for each sweep | +----------------------------------+-------------------------------------------+ | ``unit_cell=`` | Provide a target unit cell to help | | | indexing | +----------------------------------+-------------------------------------------+ | ``space_group=`` | Provide a target space group to help | | | indexing | +----------------------------------+-------------------------------------------+ | ``-atom X`` | Anomalous flag: don't merge Friedel pairs | +----------------------------------+-------------------------------------------+ Once xia2 has finished processing all the data, the final merging statistics reported by xia2 are as follows:: For AUTOMATIC/DEFAULT/SAD High resolution limit 1.29 5.63 1.29 Low resolution limit 68.02 68.02 1.33 Completeness 67.2 99.6 2.8 Multiplicity 6.5 9.2 1.1 I/sigma 16.1 27.6 4.3 Rmerge 0.124 0.136 0.000 Rmeas(I) 0.144 0.162 0.608 Rmeas(I+/-) 0.138 0.149 0.000 Rpim(I) 0.047 0.052 0.430 Rpim(I+/-) 0.058 0.060 0.000 CC half 0.988 0.956 0.823 Wilson B factor 18.807 Anomalous completeness 53.6 97.9 0.2 Anomalous multiplicity 2.8 4.9 1.0 Anomalous correlation 0.208 0.328 0.000 Anomalous slope 1.303 0.000 0.000 dF/F 0.076 dI/s(dI) 1.056 Total observations 356049 10633 187 Total unique 55180 1161 176 Assuming spacegroup: P 61 2 2 Unit cell: 92.620 92.620 128.330 90.000 90.000 120.000 These merging statistics are for all the data from all 20 sweeps merged together. Depending upon the quality of the particular datasets, and what exactly we want to do with the processed data in the next step, this may well be good enough, however in some cases we may want to analyse the individual datasets for non-isomorphism using BLEND_ with the hope of finding a subset of the data that gives a better quality merged dataset than merging the all of the data together. We can run BLEND_ in *analysis* mode on the integrated reflection files output by XDS as follows:: find `pwd` -name "INTEGRATE.HKL" > data.txt blend –a data.txt Because we are not running with keywords, press enter while it’s running then check the resulting dendrogram:: display tree.png .. image:: /figures/thermolysin_blend_tree.png If we wish we can now run BLEND_ in *synthesis* or *combination* mode to scale together and merge subsets of the data, for example to get the merged dataset for the second highest node in the dendrogram:: blend -s 10 5 pressing enter several times as before. This gives the resulting merging statistics as reported by BLEND_:: ********* Cluster 18, composed of datasets 1 2 3 5 6 7 8 9 11 13 14 15 16 18 19 20 ********* Collating multiple mtz into a single mtz ... Running AIMLESS on the unscaled file ... Statistics for this group: Rmeas Rpim Completeness Multiplicity LowRes HighRes Overall 0.133 0.074 86.4 2.6 50.15 2.48 InnerShell 0.175 0.096 85.2 2.7 50.15 8.93 OuterShell 0.171 0.096 87.9 2.6 2.58 2.48 For more information on using BLEND_ see the `BLEND tutorials`_. .. _BLEND: http://www.ccp4.ac.uk/html/blend.html .. _BLEND tutorials: http://www.ccp4.ac.uk/tutorials/tutorial_files/blend_tutorial/BLEND_tutorial.html